An advanced vocabulary causality casebook: cause, reason, mechanism, consequence, association, prediction, evidence and the question “how much causal force did that one word just add?”

The facts are small.

Maya begins using a three-step checking routine.

Her next two timed papers contain fewer sign errors.

Jia Jun writes:

The checking routine caused Maya’s sign errors to fall.

Hana changes one word:

Maya’s sign errors fell after she began using the checking routine.

Ethan changes the structure:

The fall in sign errors was associated with the period in which Maya began using the checking routine.

Same learner.

Same two papers.

Very different claim strength.

The first says the routine produced the change.

The second says the improvement came later in time.

The third says the two occurred together in the observed period without committing to cause.

This is the core of advanced causal vocabulary:

Words do not merely connect facts. They tell the reader what relationship you are claiming exists between the facts.

Because is stronger than simple sequence.

Caused is stronger than was associated with.

Led to may sound causal even when a writer only has correlation.

Predicted can describe successful forecasting without proving intervention would change the outcome.

Due to can compress an entire causal explanation into three words.

Therefore can make a conclusion sound inevitable when the premise supports only a possibility.

Cambridge Grammar treats because, as, since, so and related conjunctions as language for linking causes, reasons and results. British Council IELTS materials similarly teach expressions such as because, because of, due to, results in, as a result, causes and gives rise to. Those resources tell us how English encodes cause and result.

Research methodology adds a second question:

What kind of evidence is strong enough to justify that causal wording?

A current reference guide on statistics and research methods in the NCBI Bookshelf states the basic warning plainly: association is not causation, and causal inference from observational data usually rests on a less secure foundation than randomised experiments. Classic causal-inference work by Hernán and Robins makes the same distinction: in ideal randomised experiments, association can often be interpreted causally because randomisation creates comparability; in observational studies, association is not generally causation because the compared groups may differ in other ways.

That does not mean causal language is forbidden outside experiments.

It means causal language carries a burden.

If you write X caused Y, you owe the reader more than the fact that X and Y happened together.

You owe a credible argument about:

• temporal order;
• mechanism;
• comparison;
• alternative explanations;
• confounding;
• measurement;
• replication or converging evidence;
• the scope of the claim.

This article owns one narrow applied job in the eduKatePunggol vocabulary estate: how learners choose, read and audit causal language so the strength of the causal wording matches the strength and structure of the evidence.

It is distinct from Advanced Vocabulary Sequence | The Timeline Audit, which owns chronology and state change; Advanced Vocabulary Argument | The Qualification Audit, which owns claim strength through hedging and qualifiers; Advanced Vocabulary Reading | The Assumption Audit, which owns background commitments; and Advanced Vocabulary Perspective | The Framing Audit, which owns salience and interpretive windows.

The Timeline Audit asks:

Did B happen after A?

The Cause–Mechanism Audit asks:

What justifies saying A produced B?

The teaching examples below are constructed rather than reports of actual eduKate students or measured interventions. The external research links support the general language and causal-inference principles; the Cause–Mechanism Audit is an original classroom application.

The fifty-second route: downgrade the sentence until the evidence catches up

When you meet a causal claim, do this:

1. Underline the causal word. Because, caused, led to, resulted in, due to, therefore, consequently, produced, drove, explains, responsible for.
2. Write the two variables or events. What is the proposed cause? What is the proposed effect?
3. Check time. Did the proposed cause occur before the effect?
4. Check comparison. What would have happened without the proposed cause?
5. Check mechanism. By what plausible process could X change Y?
6. Check alternatives. Could a third factor influence both?
7. Check the design. Experiment, observational comparison, anecdote, before-after case, time series, expert judgement?
8. Match the verb to the evidence. If causal evidence is weak, downgrade to followed, coincided with, was associated with, predicted, was consistent with or another accurate form.

A powerful editing rule follows:

If changing caused to was associated with makes the sentence suddenly feel much safer, the original causal burden was doing real work.

Reader routes: Claim ladder · Because and due to · Therefore and result language · Association and prediction · Mechanism · Confounding · AI causal drift · Extended cases · Independent practice · Sources.

The causal claim ladder: from co-occurrence to intervention

Not every relationship word sits at the same level.

Use this ladder as a practical guide, not as a rigid dictionary of scientific certainty.

Level 1: co-occurrence

X and Y occurred during the same period.

No directional relation claimed.

Level 2: sequence

Y increased after X changed.

Direction in time, not cause.

Level 3: association

X was associated with Y.

A patterned relationship is observed.

Direction and cause remain open.

Level 4: prediction

X predicted Y in this dataset.

X contains information useful for forecasting Y.

This still does not mean changing X would change Y.

Level 5: contribution

X may contribute to Y.

A causal role is proposed, but the wording acknowledges that X may be one factor among several and the inference is qualified.

Level 6: mechanism-supported causation

X causes Y through mechanism M under conditions C.

Now the claim includes direction, intervention logic and a plausible process.

Level 7: quantified causal effect

Changing X by this amount changed Y by this amount in this population under these conditions.

This is the strongest and most specific form.

Notice what happens as we climb.

The language becomes more useful for intervention.

It also becomes more expensive in evidence.

That is the central exchange rate of the article:

Stronger causal language buys more explanatory power by spending more evidence.

Laboratory A: because, because of, due to, since and as — grammar can hide a stronger causal commitment

Cambridge Grammar draws a straightforward grammatical distinction:

• because introduces a clause;
• because of introduces a noun phrase;
• since and as can also introduce reasons in many contexts.

The advanced problem begins after grammar is correct.

Does the evidence actually justify a reason or cause?

Case 1. “Because” does not merely connect two true sentences

Facts:

• Maya used a new checking routine.
• Maya made fewer sign errors on the next paper.

Sentence:

Maya made fewer sign errors because she used the new checking routine.

The two component statements may both be true.

The word because adds a third claim:

the routine explains the reduction.

Possible alternatives:

• the second paper was easier;
• Maya slept better;
• she had already improved before the routine;
• the observed difference was chance variation;
• she changed several study behaviours at once.

Therefore the correct audit is not:

“Are both halves true?”

It is:

“Is the explanatory link true?”

Case 2. Because of compresses the same burden into a noun phrase

The match was cancelled because of the rain.

Cambridge treats because of as a preposition meaning roughly “as a result of”.

In everyday contexts, the cause may be obvious and well established.

Now educational writing:

The student’s score fell because of weak vocabulary.

What does weak vocabulary mean?

• unknown words?
• slow retrieval?
• misread command words?
• poor collocation?
• insufficient subject knowledge?
• weak reading inference that merely looks like vocabulary?

The grammar is correct.

The causal diagnosis is under-specified.

Better:

The student lost six marks after misreading three command words. Those language errors directly changed what she attempted in the questions.

Now the mechanism becomes visible.

Case 3. Due to can sound more objective than because without becoming more evidential

The delay was due to traffic.

The delay happened because there was heavy traffic.

The first sounds more compact and formal.

It does not magically provide stronger evidence.

This matters in reports:

The decline was due to reduced motivation.

The noun phrase can sound diagnostic while hiding how motivation was measured.

Ask:

• Who measured motivation?
• How?
• When?
• What alternative causes were considered?

Formal register does not upgrade causal evidence.

Case 4. “Since” can quietly switch from time to reason

Timeline sentence:

Attendance improved since the timetable changed.

Possible reading:

attendance improved from that time onward.

Reason reading:

Since the timetable was easier for families, attendance improved.

Now since means approximately because.

Where both readings are possible, causal analysis deserves cleaner language:

Attendance improved after the timetable changed.

for chronology.

Attendance improved because the new timetable reduced a transport conflict.

for causation—if supported.

Case 5. “As” can hide whether the writer means time or reason

As Maya became tired, her error rate rose.

Does as mean:

• while fatigue increased?
• because fatigue increased?

The sentence may invite both.

Better temporal version:

Maya’s error rate rose during the period in which she reported increasing fatigue.

Better causal version, if evidence supports it:

Fatigue contributed to the rise in Maya’s error rate by reducing her end-of-paper checking.

The second gives a mechanism.

The reason-clause audit

1. Is the sentence merely chronological?
2. Does because/due to add an explanation not separately established?
3. Is the proposed cause measured or just labelled?
4. Would after or was associated with be more accurate?
5. Can the writer name a mechanism rather than a vague cause noun?

Laboratory B: therefore, consequently, thus, so, leads to and results in — conclusion words can outrun the premise

Cambridge and British Council materials teach a useful family of cause-result expressions:

• so;
• therefore;
• consequently;
• as a result;
• thus;
• results in;
• leads to;
• causes;
• gives rise to.

The vocabulary is easy to memorise.

The logic is harder.

Case 6. Therefore can make a weak inference sound finished

Premise:

Students who read more tended to have larger vocabularies.

Conclusion:

Therefore, increasing reading time will enlarge every student’s vocabulary.

The conclusion is stronger than the premise.

Possible reasons:

• students with larger vocabularies may enjoy reading more;
• family literacy environment may influence both;
• reading quality may matter more than time alone;
• the relationship may differ across learners.

Therefore itself is not the error.

The unsupported inferential jump is.

Safer:

The association is consistent with the possibility that sustained reading contributes to vocabulary growth, but the observed relationship alone does not show how much additional reading would cause vocabulary to increase.

Case 7. “Leads to” sounds softer than “causes” but often carries the same direction

Sleep deprivation leads to poor examination performance.

This is causal language.

It presents sleep deprivation as producing poorer performance.

If the evidence only shows that students sleeping less tend to score lower, write:

Shorter sleep duration was associated with lower scores in the observed group.

If an experiment or strong converging evidence supports causal interpretation, leads to may be appropriate.

Do not treat softer rhythm as weaker logic.

Case 8. Results in presents a directional production relationship

Misreading the word except results in selecting the wrong option.

In a tightly constructed item where the misreading directly changes the selection logic, the causal claim can be strong.

The mechanism is immediate:

the learner reverses the inclusion rule.

Compare:

Weak vocabulary results in low examination scores.

Too broad.

The cause category is vague, the outcome has many causes, and the mechanism is unspecified.

Stronger:

When a learner misreads a command word, the misunderstanding can cause an otherwise knowledgeable response to answer the wrong task.

Specific cause.

Specific mechanism.

Specific failure mode.

Case 9. Consequently may be formally elegant and logically wrong

The class received a new workbook. Consequently, average scores rose.

The sentence structure suggests the workbook caused the rise.

But perhaps:

• the next test was easier;
• students had six additional weeks of schooling;
• a different topic was tested;
• the class composition changed.

Formal connectors can make a paragraph sound academically mature while making the reasoning less disciplined.

Academic style is not the density of therefore.

It is the quality of the inferential bridge.

Case 10. “So” can be causal, inferential or conversationally loose

It rained, so the match was cancelled.

Reason-result relation.

The lights are off, so they must have gone home.

Now so supports an inference, not a physical cause.

The lights being off did not cause the people to leave.

They are evidence from which the speaker infers departure.

This distinction matters in comprehension:

cause-result language and evidence-conclusion language can share the same connector.

The result-word audit

1. Is the second clause a result in the world or a conclusion in the writer’s reasoning?
2. Does the premise justify the strength of therefore/consequently?
3. Does leads to/results in imply causation?
4. Would followed by or was associated with be more accurate?
5. Is the writer using formal vocabulary to hide a weak inferential bridge?

Laboratory C: association, correlation, prediction and causation — four relationships learners must not collapse

Scientific and analytical reading becomes much easier when four ideas are kept separate.

Association

X and Y vary together in some observed way.

Example:

Higher attendance was associated with higher test scores.

Possible explanations:

• attendance improves learning;
• more motivated students both attend and study more;
• health affects both attendance and scores;
• family support affects both;
• some combination of the above.

The association is real if measured well.

The cause remains open.

Correlation

Correlation is a particular statistical description of co-variation, often used for numerical variables.

Everyday writing often uses correlated with loosely.

Advanced writing should reserve it for an actual measured correlation when possible.

Do not write:

Vocabulary is correlated with better writing

if no statistical relationship was calculated or cited.

Use is related to or a more exact description of the evidence.

Prediction

A predictor helps forecast an outcome.

That does not mean intervening on the predictor will change the outcome.

Example:

Suppose shoe size predicts reading ability in a mixed sample of young children.

Larger shoes may predict better reading because older children tend to have both larger feet and more developed reading skills.

Increasing shoe size would not teach reading.

Prediction can be useful without being causal.

Causation

Causal language claims that changing X would make a difference to Y, all else appropriately considered.

This is why causal claims are naturally connected to intervention:

If X really causes Y, changing X should alter Y under the relevant conditions.

That idea is more demanding than prediction.

A weather forecast can predict rain without causing it.

A thermometer can predict fever severity without causing the fever.

An examination diagnostic can predict who will struggle without being the cause of the struggle.

Case 11. “Associated with” is not a coward’s version of “causes”

Sometimes students think cautious language sounds weak.

It can be more precise.

If the study design establishes association but not causation, associated with is not evasive.

It correctly names the result.

The error is either:

• upgrading association into causation without support;
• or refusing causal language even when the design and reasoning genuinely support it.

Precision works in both directions.

Case 12. “Predicts” can be strong evidence for forecasting and weak evidence for intervention

Early vocabulary score predicted later reading performance.

This may be valuable.

It can help identify learners needing support.

But it does not by itself prove that simply increasing the measured vocabulary score would produce the same later reading gains.

The predictor may be standing in for broader language development, reading exposure, world knowledge or home literacy environment.

Prediction answers:

Who is likely to show Y?

Causation answers:

What would happen to Y if we changed X?

Those are related questions.

They are not the same question.

The relationship audit

1. Was a statistical relationship measured? If not, avoid casual use of correlated.
2. Does X help forecast Y? That is prediction.
3. Does X merely co-vary with Y? That is association.
4. Would changing X change Y? That is the causal question.
5. Has the study design earned the causal answer?

Laboratory D: mechanism — causal language becomes stronger when the middle is visible

A weak causal sentence often has only two boxes:

X → Y.

A stronger explanation opens the arrow.

X → mechanism → intermediate change → Y.

Mechanism does not prove causation by itself.

But a plausible, observed mechanism makes a causal claim easier to test and harder to fake.

Case 13. “Vocabulary caused the improvement” is too far away from the action

Weak:

Better vocabulary caused Maya’s comprehension score to rise.

What happened in between?

Possible mechanism:

• Maya learned the meaning of reluctant;
• she recognised it in a passage;
• that recognition changed her interpretation of a character’s decision;
• the revised interpretation supported the correct inference answer.

Now the claim can become:

Learning the meaning of reluctant allowed Maya to interpret the character’s hesitation correctly, which in turn supported the correct inference answer.

The mechanism is local and inspectable.

It does not claim that vocabulary explains the whole paper.

Case 14. A mechanism can be plausible and still wrong

Suppose a tutor says:

The new highlighting routine improved comprehension because it directed attention to key evidence.

Plausible.

But perhaps the learner simply slowed down because the new routine forced slower reading.

Or perhaps the passage was easier.

Or perhaps the learner had already become more familiar with the question type.

A mechanism is a testable story, not an automatic proof.

Ask:

• Did attention actually change?
• Did the learner select more relevant evidence?
• Did time-on-task change?
• Did the benefit persist on a different passage?

Case 15. Mechanism can rescue an otherwise vague educational label

Weak:

The student performed poorly because of low confidence.

What does confidence do?

Stronger mechanism hypothesis:

When uncertain, the student abandoned correct first answers and changed them without new evidence. This answer-switching behaviour accounted for four lost marks in the observed paper.

Now the tutor can intervene on behaviour rather than personality.

The cause claim also becomes narrower:

uncertainty-triggered answer switching contributed to specific lost marks.

Case 16. Mechanism and mediation are not the same as confounding

Suppose:

reading practice → vocabulary growth → better comprehension.

If vocabulary growth is one path through which reading practice affects comprehension, vocabulary is acting as a mediator in the proposed chain.

Now suppose:

family literacy environment → more reading practice

and:

family literacy environment → stronger comprehension.

The family literacy environment may confound the observed relationship between reading practice and comprehension.

One sits on the causal pathway.

The other may create a non-causal or exaggerated association between the variables being compared.

Advanced reading should not treat every third variable as the same kind of third variable.

The mechanism audit

1. What happens between X and Y?
2. Can the intermediate steps be observed?
3. Would the proposed mechanism predict additional evidence?
4. Could another mechanism produce the same outcome?
5. Is the writer confusing a mediator with a confounder?

Laboratory E: confounding, reverse causation and common causes — three reasons an association can mislead

The NCBI reference materials on causal inference repeatedly stress that observational associations may not represent causal effects because the compared groups can differ in other important ways.

For classroom purposes, three failure patterns are enough to transform how students read claims.

Pattern 1. Common cause

Observed:

Students who own more books tend to score higher in reading.

Tempting claim:

Owning more books causes higher reading scores.

Possible common causes:

• family reading habits;
• parent education;
• household resources;
• child interest in reading;
• access to libraries and enrichment.

Buying books may still help.

The raw association alone does not identify the effect of the purchase.

Pattern 2. Reverse causation

Observed:

Students who use dictionaries more often have weaker vocabulary scores.

Bad conclusion:

Dictionary use weakens vocabulary.

Reverse explanation:

students with weaker vocabulary may need dictionaries more often.

The direction could be:

weaker vocabulary → more dictionary use.

Or both directions may operate.

Direction cannot be read from association alone.

Pattern 3. Selection effect

Observed:

Students attending extra classes have lower average scores than students who do not.

Bad conclusion:

Extra classes cause lower scores.

Alternative:

students who are already struggling may be more likely to seek extra classes.

The treatment group began from a different state.

This is a classic reason ordinary group comparisons can mislead.

Case 17. The tuition paradox

Imagine a school survey finds:

students receiving tuition score 5 points lower on average than students without tuition.

Possible explanation A:

tuition is ineffective or harmful.

Possible explanation B:

students were sent to tuition because their scores were already lower.

Possible explanation C:

the tuition group contains harder cases, while the non-tuition group contains more independent learners.

Possible explanation D:

some tuition helps, some does not, and average group comparison hides that variation.

The observed 5-point difference is real data if measured correctly.

Its causal interpretation is not automatic.

Case 18. Regression to the mean can mimic improvement

A student takes an unusually bad test.

The family begins a new intervention immediately.

The next score rises.

Tempting conclusion:

The intervention caused the recovery.

But extremely low scores are often followed by less extreme scores even without an intervention because random bad luck does not repeat perfectly.

The before-after pattern can therefore exaggerate apparent treatment effects.

This is another reason one dramatic recovery is evidence worth investigating, not necessarily a complete causal proof.

Case 19. Simultaneous change creates causal ambiguity

Suppose Maya changes four things before the next examination:

• new checking routine;
• earlier bedtime;
• more timed papers;
• less phone use during revision.

Her score rises.

Which change caused how much?

The package may have helped.

Attributing the full improvement to one favourite component is not justified without additional evidence.

This is why causal analysis asks not merely “what changed?” but “what changed separately enough to learn from?”

The alternative-explanation audit

1. Could Y influence X instead?
2. Could a third variable influence both?
3. Were people selected into groups because of their starting state?
4. Did several things change at once?
5. Could the apparent improvement partly reflect regression to the mean?

Necessary causes, sufficient causes and contributing causes

Everyday language often treats “the cause” as singular.

Complex outcomes rarely cooperate.

Necessary cause

If X is necessary for Y, Y cannot happen without X.

Educational examples of truly necessary causes are less common than students think.

For a particular question, understanding the instruction may be necessary for answering that exact instruction correctly.

But “knowing 5,000 words is necessary for good writing” is a much larger claim that would need careful definition and evidence.

Sufficient cause

If X is sufficient for Y, X by itself is enough to produce Y under the stated conditions.

Again, complex educational outcomes rarely have one simple sufficient cause.

Knowing a word is not sufficient for high comprehension if syntax, background knowledge or inference fails.

Contributing cause

This is often the most realistic educational language.

Vocabulary knowledge contributed to comprehension by reducing uncertainty around several key sentences.

The sentence claims a causal role without pretending vocabulary was the only cause.

Words such as contributed to, partly explains, one factor was, may have increased can be excellent when they match the multi-causal structure.

But qualification must not become decoration.

May have contributed is still a causal suggestion.

The writer still owes evidence.

The cause-type audit

1. Is the writer claiming X is necessary?
2. Is the writer claiming X is sufficient?
3. Would contributing cause be more realistic?
4. Is the outcome obviously multi-causal?
5. Has “the cause” erased other active factors?

The evidence-design ladder: what kind of study earns what kind of wording?

A causal verb should not be chosen only from a thesaurus.

It should be chosen partly from the study design.

The ladder below is deliberately simplified for learners. Real causal inference can be more sophisticated than a one-dimensional hierarchy, and strong observational designs can sometimes support serious causal conclusions. The purpose here is to teach evidence matching, not to turn one research design into a magic stamp.

Evidence form 1: anecdote

I used the new routine and my score rose.

Useful for:

• generating a hypothesis;
• identifying a possible mechanism;
• deciding what to test next.

Weak for:

• generalising to everyone;
• separating intervention effects from coincidence;
• quantifying causal effect.

Good wording:

The improvement followed the introduction of the routine and is consistent with the possibility that the routine helped.

Evidence form 2: repeated within-person pattern

Suppose Maya repeatedly makes fewer sign errors when she uses the checklist and more when she omits it across several comparable papers.

This is stronger than one anecdote because the pattern repeats within the same learner.

But comparability still matters:

• Were the papers equally difficult?
• Did other study behaviours change?
• Was time pressure similar?

Good wording:

Across the observed papers, checklist use repeatedly coincided with fewer sign errors, supporting the hypothesis that the checklist contributes to error reduction.

Evidence form 3: before–after group comparison

A class takes a pre-test, receives an intervention and improves on a post-test.

Stronger than one anecdote.

Still vulnerable to:

• practice effects;
• maturation;
• different test difficulty;
• other teaching received during the period;
• regression to the mean.

Good wording:

Scores improved after the intervention, but the before–after design alone does not isolate how much of the change was caused by the intervention.

Evidence form 4: observational comparison with measured covariates

Researchers compare students who received an intervention with those who did not and adjust for measured differences such as prior score or age.

This can substantially improve causal inference.

But unmeasured confounding may remain.

Good wording depends on the design quality and analytical assumptions.

Possible:

After adjustment for prior score and age, intervention participation remained associated with higher later scores.

If a formal causal method and assumptions justify it, stronger causal language may be defensible—but the assumptions should be visible.

Evidence form 5: natural experiment or quasi-experimental design

Sometimes external circumstances create groups that are more comparable than ordinary self-selection would allow.

Current statistical reference materials describe natural or quasi-experiments as situations in which an external change can create treatment and control groups in a way that resembles random assignment more closely than ordinary observation.

Such designs can support stronger causal reasoning when their assumptions hold.

Evidence form 6: randomised experiment

Participants are randomly assigned to conditions.

Randomisation aims to make the groups exchangeable on average so systematic confounding is reduced.

This is why ideal randomised experiments support causal interpretation much more directly.

Even then, ask:

• Was randomisation implemented properly?
• Did participants comply?
• Was there attrition?
• Was the outcome measured fairly?
• Does the result generalise beyond the sampled population?

Randomised does not mean flawless.

It means a major source of causal ambiguity has been deliberately attacked.

Evidence form 7: converging evidence

Strong causal knowledge often comes not from one perfect study but from several different kinds of evidence pointing in the same direction:

• randomised trials;
• natural experiments;
• mechanism evidence;
• dose-response patterns;
• replication across settings;
• time ordering;
• well-designed observational analysis.

If different methods with different weaknesses converge, confidence can rise.

This is why advanced causal reading should not reduce every question to “Was it randomised?”

Ask instead:

What alternative explanation did this design rule out, and what alternative explanations remain?

The design-to-language audit

1. Anecdote? Prefer hypothesis language.
2. Before–after only? Preserve chronology and uncertainty.
3. Observational comparison? Name association unless causal assumptions are explicitly defended.
4. Quasi-experimental or natural experiment? Inspect the assignment mechanism.
5. Randomised experiment? Causal language may be justified within the tested population and conditions.
6. Multiple converging designs? Evaluate the whole evidence pattern rather than worshipping one label.

Cause versus reason: human explanations can be different from physical causes

Language complicates causality because because can introduce different kinds of explanation.

Compare:

The glass broke because it fell onto the floor.

Physical causal explanation.

Maya left because she wanted to catch the bus.

Reason or motive explanation.

Hana must be home because her shoes are at the door.

Evidence-to-conclusion explanation.

The same connector supports three structures:

• event causes event;
• reason motivates action;
• evidence supports inference.

Advanced comprehension depends on identifying which one is active.

Case 20. Motive is not always the same as cause

Jia Jun apologised because he wanted the argument to end.

The sentence gives a reason for his decision.

But the full cause of the apology may include:

• parent pressure;
• guilt;
• social consequences;
• desire to restore the friendship.

A stated motive is one explanation offered by or about an agent.

It may not be the only cause of the behaviour.

This matters in narrative analysis because characters’ stated reasons can be incomplete, strategic or self-deceptive.

Case 21. Evidence can justify a conclusion without causing the underlying event

The pavement is wet, so it probably rained.

The wet pavement does not cause the rain.

It is evidence used to infer rain.

Another possible cause exists:

a sprinkler.

This tiny example contains the whole discipline:

• event relation;
• evidence relation;
• alternative explanation;
• claim strength.

The three-because test

Whenever you see because, ask which one it is:

1. Cause because: X produced Y.
2. Motive because: an agent acted for reason R.
3. Evidence because: fact E supports conclusion C.

If the learner cannot classify the because, they probably have not yet understood the explanatory relationship.

Cross-subject causal language: the arrow changes meaning across disciplines

English: character motive, plot consequence and evidence inference

A literature answer may contain several causal levels at once:

Maya hides the letter because she fears Jia Jun’s reaction. This decision causes Hana to believe the letter is missing, which later contributes to the conflict.

Reason:

fear motivates hiding.

Plot cause:

hiding changes Hana’s information state.

Contributing cause:

the mistaken belief becomes one factor in later conflict.

Strong analysis does not use causes for every relationship.

Mathematics: proof is not empirical causation

Because two angles form a straight line, their sum is 180°.

This is not an observational causal claim.

The relationship follows from mathematical definitions and properties.

Therefore in a proof marks logical consequence.

It does not mean one angle physically causes the other angle to have a certain measure.

Students who learn “therefore = result” too mechanically can confuse logical entailment with real-world cause.

Science: causal mechanism is often the point

Science often asks not only whether variables move together but how one process changes another.

Example:

Increasing temperature increased reaction rate because particles collided more frequently with sufficient energy under the tested conditions.

Now the explanation contains:

• intervention: temperature increase;
• outcome: reaction rate increase;
• mechanism: altered collision behaviour;
• scope: tested conditions.

The language is causal because the knowledge claim is causal.

Humanities: causes usually form systems, not single arrows

Historical questions often invite monocausal writing:

The war happened because of X.

Stronger historical reasoning may separate:

• long-term structural conditions;
• short-term pressures;
• triggering events;
• decision-makers’ motives;
• contingent events.

Vocabulary such as contributed to, intensified, enabled, constrained, triggered, accelerated, made more likely can represent different causal jobs more accurately than repeating caused.

Family life: explanations should diagnose the controllable link

Weak family explanation:

You were late because you are irresponsible.

Trait label.

Low intervention value.

Mechanism version:

You started packing only after the planned departure time, so the family left twelve minutes late.

Observable cause.

Controllable next step.

Good causal language can reduce moral heat because it moves from identity to mechanism.

Laboratory F: AI causal drift — when clearer prose becomes a stronger claim

AI rewriting has a predictable temptation.

Writers ask:

Make this clearer.

Make this more concise.

Make this more persuasive.

Make this sound more academic.

A model may then replace cautious observational language with cleaner causal language.

The sentence improves stylistically.

The evidence does not.

Case 22. “After” becomes “because”

Source:

Scores rose after the revision routine was introduced.

AI rewrite:

Scores rose because the revision routine was introduced.

The first sentence is temporal.

The second is explanatory.

Nothing in the rewrite process earned the new explanation.

Repair:

Scores rose after the revision routine was introduced, but the sequence alone does not establish that the routine caused the increase.

Case 23. “Associated with” becomes “drives”

Source:

Higher reading frequency was associated with larger vocabulary scores.

AI marketing rewrite:

Frequent reading drives vocabulary growth.

Drives is vivid.

It is also causal.

It suggests intervention on reading frequency will produce growth.

The source may support that hypothesis.

It does not automatically prove it.

Case 24. “Predicted” becomes “determined”

Source:

Early vocabulary score predicted later reading performance.

AI rewrite:

Early vocabulary determines later reading performance.

This is much stronger.

Determines can suggest necessity, sufficiency or near inevitability depending on context.

Prediction has become destiny.

That is not editing.

It is theory invention.

Case 25. “May contribute to” becomes “leads to”

Source:

Sleep loss may contribute to end-of-paper checking errors.

AI concise rewrite:

Sleep loss leads to checking errors.

Two changes occurred:

• may disappeared;
• contribute to became leads to.

The sentence moved from qualified multi-causality to direct causation.

The shorter version is not a paraphrase.

It is a stronger claim.

Case 26. “Therefore” appears where the source said “possibly”

Source:

The two groups differed in attendance. This difference may partly explain the score gap.

AI rewrite:

The two groups differed in attendance; therefore, attendance explains the score gap.

The model removed:

• uncertainty;
• partial contribution;
• alternative explanations.

A semicolon cannot repair an inference.

Case 27. AI neutralisation can erase a justified causal claim too

Causal drift can move downward as well.

Suppose a carefully randomised experiment finds that a specific instructional intervention produces a measured improvement under the tested conditions.

Source:

Random assignment to the retrieval-practice condition increased delayed recall relative to the control condition.

AI “safe” rewrite:

Retrieval practice was associated with higher delayed recall.

The rewrite may be technically true but unnecessarily discards causal information earned by the design.

Good evidence deserves precise strength too.

The goal is not permanent hedging.

It is calibrated language.

A practical AI prompt for causal-strength preservation

Use:

Rewrite for clarity without changing causal strength. Preserve the distinction among sequence, association, correlation, prediction, contribution and causation. Do not replace after with because, associated with with causes/drives/leads to, predicted with determines, or may contribute to with direct causal language unless the source explicitly supports the stronger claim. Do not weaken an experimentally justified causal statement into mere association without explaining the loss. Flag every change to because, due to, therefore, consequently, results in, leads to, causes, contributes to, associated with, correlates with, predicts, determines, explains and responsible for.

Then run one manual test:

Could a reasonable reader take a different action because of the stronger wording?

If yes, the causal change is material.

The intervention test: the fastest way to expose hidden causal meaning

Take a sentence and imagine acting on it.

Sentence:

Students who sleep longer score higher.

Action question:

If we make a student sleep one additional hour, should we expect the score to rise?

The original sentence does not answer that.

Now causal sentence:

Increasing sleep duration improves examination performance.

This sentence does invite the intervention.

That is why causal language matters.

It does not merely explain the past.

It tells us what might happen if we change the world.

The counterfactual question: what would have happened otherwise?

A causal claim compares the observed world with an unobserved alternative.

Maya used the new routine and scored 80.

To know the causal effect of the routine on Maya, we want to know:

What would Maya have scored on the same occasion, under the same conditions, without the routine?

We cannot observe both worlds at once.

Research design tries to build a credible substitute for the missing counterfactual.

A comparison group.

A randomised control condition.

A repeated baseline.

A natural experiment.

A statistical adjustment under explicit assumptions.

This is the deep reason “before and after” is not automatically causal.

We saw what happened after.

We did not see what would have happened instead.

A family-scale version of the counterfactual

Parent:

Your marks improved because I took away your phone.

Possible.

But perhaps the child also:

• started revision earlier;
• received school feedback;
• changed sleep habits;
• had an easier paper;
• finally understood a topic that had been unstable for weeks.

A better family question is:

What changed after the phone rule, what mechanism could connect that change to the marks, and what other changes happened during the same period?

This does not require a home randomised trial.

It requires intellectual humility.

Families can make sensible decisions without pretending every improvement has one proven cause.

Extended causal case files: build the explanation, then attack it

The hardest causal errors are not obvious falsehoods.

They are plausible explanations that became stronger than the evidence while nobody was watching the verb.

Case file 1. The checking routine that “worked”

Constructed sequence.

• Paper 1: Maya makes six sign errors.
• A tutor introduces a three-step checking routine.
• Paper 2: Maya makes three sign errors.
• Paper 3: Maya makes two sign errors.

Claim A:

The checking routine caused Maya’s sign errors to fall.

Evidence for:

• correct temporal order;
• repeated improvement;
• plausible mechanism;
• error type targeted directly by the routine.

Evidence against complete certainty:

• papers may differ in difficulty;
• Maya may be learning generally over time;
• other revision habits may have changed;
• three observations are a small sample.

Best classroom conclusion:

The repeated reduction in sign errors after checklist introduction, together with the routine’s direct error-checking mechanism, supports the hypothesis that the routine contributed to Maya’s improvement. More comparable observations would strengthen the causal conclusion.

Notice the wording.

Not “we know nothing”.

Not “proved”.

Calibrated causal confidence.

Case file 2. The extra-class comparison

Constructed survey.

Students taking extra classes average 68.

Students not taking extra classes average 75.

Headline:

Extra classes reduce academic performance.

Immediate audit:

• Were weaker students more likely to enrol?
• What were prior scores?
• Were subjects comparable?
• Were tuition hours replacing sleep or independent practice?
• Were the extra classes all the same quality?

Better headline:

Students taking extra classes had lower average scores in this sample; the comparison does not show whether the classes caused the difference.

If prior scores show that the tuition group began ten points lower and later narrowed the gap, the causal story may reverse completely.

Group difference is not treatment effect.

Case file 3. The homework phone rule

Constructed family observation.

Before the rule, Ethan checks his phone frequently during homework.

After the rule, the phone stays outside the room.

Homework completion time falls from 95 minutes to 70 minutes across several evenings.

Possible mechanism:

• fewer interruptions;
• less task-switching;
• longer uninterrupted work periods.

Alternative explanations:

• homework was easier that week;
• Ethan had become familiar with the topic;
• school workload changed;
• bedtime improved too.

Family decision:

It can still be reasonable to keep the phone rule because the mechanism is plausible, the pattern repeated and the cost is acceptable.

Scientific certainty is not required for every household decision.

But the family should distinguish:

This rule appears to help Ethan work more continuously

from:

Phones cause poor academic performance.

One is a local decision claim.

The other is a broad causal generalisation.

Case file 4. The reading association

Constructed dataset.

Students reporting more weekly reading also tend to score higher in vocabulary.

Possible causal direction:

more reading → more word encounters → vocabulary growth.

Possible reverse direction:

larger vocabulary → easier reading → more voluntary reading.

Possible common causes:

• family literacy environment;
• school achievement;
• interest in language;
• access to books;
• parent modelling.

The relationship may be genuinely bidirectional:

reading improves vocabulary, and vocabulary makes reading easier, producing a feedback loop.

This is a good example of why “X causes Y” can be too small for a system.

Sometimes the right model is:

X ↔ Y within a wider environment.

Case file 5. The examination intervention

Constructed trial.

Forty students are randomly assigned to two groups.

One group receives a ten-minute command-word training sequence.

The other group receives an equal-length unrelated review activity.

Both then complete a new set of questions.

The command-word group makes fewer task-selection errors.

Causal wording becomes more defensible:

Under the tested conditions, assignment to the command-word training reduced task-selection errors relative to the control activity.

Notice the limits remain visible:

• under the tested conditions;
• relative to this control;
• on this outcome;
• in this sample.

Strong causal language can still be scoped.

Case file 6. The science experiment where mechanism and manipulation align

Constructed experiment.

Students vary one factor while holding relevant conditions stable.

As temperature rises across the tested range, reaction time falls in a consistent pattern.

A known mechanism explains why increased kinetic energy changes collision frequency and effective collisions.

Causal claim:

Increasing temperature increased reaction rate under the tested conditions.

This is stronger than:

Temperature was associated with reaction rate.

The stronger sentence is justified because temperature was manipulated, other relevant conditions were controlled and the mechanism is coherent with the observed pattern.

Case file 7. The “motivation” diagnosis

Constructed learner pattern.

Jia Jun completes easy homework quickly but delays starting difficult unfamiliar tasks.

Adult explanation:

He delays because he lacks motivation.

Alternative mechanisms:

• task instructions are unclear;
• he cannot identify the first step;
• he expects failure;
• he is mentally fatigued;
• he uses avoidance to escape uncertainty.

Diagnostic experiment:

Give only the first-step cue without solving the problem.

If Jia Jun then works independently for twenty minutes, the original “low motivation” explanation becomes less attractive.

The behaviour may have been generated by a start-state bottleneck rather than a global motivational deficit.

This is causal reasoning as teaching.

Change one suspected cause.

Observe the downstream state.

Case file 8. The AI summary that invents a mechanism

Source notes.

• Attendance rose after the schedule changed.
• Parents reported fewer transport conflicts.
• No formal mediation analysis was conducted.

AI summary:

The schedule change improved attendance by reducing transport conflicts.

This is plausible.

It may even be correct.

But the model has joined three observations into a tested mechanism that the source did not establish.

Faithful version:

Attendance rose after the schedule change, while parents also reported fewer transport conflicts. The reports suggest one plausible mechanism for the improvement, but the available evidence does not isolate that pathway.

Longer.

Better.

The full Cause–Mechanism Audit

1. Relationship word: what causal or inferential word is doing the work?
2. Cause candidate: what exactly is X?
3. Outcome: what exactly is Y?
4. Timeline: does X precede Y?
5. Mechanism: how could X change Y?
6. Counterfactual: what would likely happen without X?
7. Alternative direction: could Y influence X?
8. Common cause: could Z influence both?
9. Selection: were groups different before exposure or treatment?
10. Concurrent change: did other things change at the same time?
11. Measurement: are X and Y actually measured, or merely labelled?
12. Design: what study structure supports the claim?
13. Scope: which population, context and outcome does the evidence cover?
14. Verb calibration: should the sentence say followed, associated with, predicted, contributed to or caused?

Independent practice: thirty causal decisions

For each item, identify the strongest relationship the evidence actually earns. Do not choose the most impressive verb. Choose the most accurate one.

Use this five-step routine:

1. Name X and Y.
2. Identify the relationship word.
3. Ask what evidence exists.
4. Generate at least one alternative explanation.
5. Rewrite the claim at the strength the evidence earns.

Practice set A: because, due to, therefore and causal verbs

1. Facts: a checking routine begins; the next paper contains fewer sign errors. Claim: The routine caused the reduction. What additional evidence would strengthen the causal wording?

2. The student’s score fell because of weak vocabulary. What is wrong with treating weak vocabulary as a sufficient diagnosis?

3. The decline was due to reduced motivation. What does the phrase hide unless motivation was actually measured?

4. Students who read more had larger vocabularies. Therefore, increasing reading time will increase every student’s vocabulary. Identify the inferential jump.

5. Sleep deprivation leads to poor examination performance. If the evidence is a cross-sectional survey only, rewrite more cautiously.

6. Misreading except results in selecting the wrong option. Under what tightly defined circumstances can this causal wording be justified?

7. The class received a new workbook. Consequently, average scores rose. Why is the connector not enough to establish the relationship?

8. The lights are off, so the family must have gone out. Is so marking physical causation or evidence-to-conclusion inference?

Practice set B: association, correlation, prediction and causation

9. Higher attendance was associated with higher scores. Name two possible explanations besides “attendance causes higher scores”.

10. A writer says vocabulary is correlated with better writing but cites no calculated correlation. What wording is safer?

11. Shoe size predicts reading ability among young children. Why would increasing shoe size not teach reading?

12. Early vocabulary score predicts later reading performance. What does that prediction fail to prove about vocabulary intervention?

13. X predicts Y and X causes Y. Which sentence invites the intervention question “what happens if we change X?”

14. A randomised experiment finds a difference between conditions. Is it always better to rewrite the result as was associated with?

Practice set C: mechanism and alternative explanations

15. Better vocabulary improved comprehension. Insert a plausible mechanism between the two concepts.

16. The highlighting routine improved comprehension because it directed attention to evidence. Name one alternative mechanism that could produce the same observed improvement.

17. Students who use dictionaries more often have weaker vocabulary scores. Give a reverse-causation explanation.

18. Students with more books at home score higher in reading. Give one plausible common cause.

19. Students attending tuition score lower than students without tuition. What selection effect might explain the association?

20. A student has an unusually bad paper, begins a new intervention, then improves. Why can regression to the mean exaggerate the apparent intervention effect?

21. Maya simultaneously changes sleep, phone use, checking routine and practice volume, then improves. Why can you not assign the whole gain to the checking routine?

Practice set D: study design and causal language

22. One student says a strategy helped after using it once. What kind of evidence is this, and what wording is appropriate?

23. A class improves from pre-test to post-test after an intervention. Name two non-intervention explanations the design does not automatically rule out.

24. An observational study adjusts for prior score and age. What important causal problem may still remain?

25. Why does random assignment strengthen a causal interpretation?

26. Why can several different methods pointing to the same mechanism be stronger than one impressive study alone?

Practice set E: AI causal drift

27. Source: Scores rose after the routine was introduced. AI: The routine caused scores to rise. What did the model add?

28. Source: Reading frequency was associated with vocabulary score. AI: Reading drives vocabulary growth. What changed?

29. Source: Early vocabulary predicted later reading. AI: Early vocabulary determines later reading. Why is the second sentence much stronger?

30. Source: a well-run randomised experiment shows that the intervention increased delayed recall. AI rewrites increased as was associated with. What useful information may be lost?

Answer commentary: match the arrow to the evidence

1. Stronger evidence could include repeated comparable papers, evidence that the routine was actually used, a mechanism linking the checks to the specific prevented errors, and some way to rule out competing changes. One before–after observation is suggestive, not decisive.

2. Weak vocabulary is a broad label rather than a mechanism. The actual problem might be unknown words, slow retrieval, command-word misreading, background knowledge or inference. A causal diagnosis should identify the operational bottleneck.

3. It hides whether motivation was measured, inferred from behaviour or simply used as a retrospective explanation. Formal noun phrases can make guesses sound like findings.

4. The premise describes an association. The conclusion claims that intervening on reading time will produce vocabulary growth in every student. Reverse causation, common causes and learner variation remain possible.

5. Example: In the surveyed group, shorter sleep duration was associated with lower examination scores. The observational relationship can be reported without claiming that sleep loss produced the score difference.

6. When the question structure makes the misreading directly reverse or alter the required selection rule, and the learner’s reasoning shows that this specific misinterpretation produced the wrong choice, the local causal chain is defensible.

7. Consequently asserts or strongly presents a result relationship, but it does not rule out easier testing, maturation, other teaching or changing class composition. A connector cannot manufacture a comparison group.

8. Evidence-to-conclusion inference. The dark house is evidence for the speaker’s conclusion. It did not physically cause the family to leave.

9. Examples: more motivated students may both attend more and study more; health may affect both absence and performance; family support may influence both attendance and achievement. The measured association does not choose among these explanations.

10. Use is related to, is associated with only if an association is actually evidenced, or simply describe the observed pattern. Correlated with should ideally refer to a measured statistical correlation.

11. Age is a common cause: older children tend to have larger feet and stronger reading. Shoe size can forecast reading in the sample without being an instructional lever.

12. It does not prove that raising the measured vocabulary score by intervention would cause the later reading improvement. The score may be a marker of broader language development or literacy environment.

13. X causes Y. Causal language naturally supports intervention reasoning: if we deliberately change X, Y should change under the relevant conditions.

14. No. If randomisation and implementation justify a causal comparison, reducing the result to mere association discards information the design legitimately earned. Precision requires neither automatic inflation nor automatic hedging.

15. Example: vocabulary knowledge reduces uncertainty around a key word, which changes sentence interpretation, which supports a more accurate inference. The middle steps make the proposed arrow inspectable.

16. The highlighting routine may simply force slower reading, and the extra time rather than highlighted evidence may improve comprehension. A plausible mechanism must compete with plausible alternatives.

17. Weaker vocabulary may cause greater dictionary use because those learners encounter more unknown words. The observed direction can therefore run from vocabulary difficulty to lookup frequency.

18. Family literacy environment is one possibility: homes with strong reading habits may both contain more books and support reading development. Household resources or parent education could also contribute.

19. Students who are already struggling may be more likely to seek tuition. The lower average can reflect who entered the treatment group rather than an adverse effect of treatment.

20. Extreme poor performance often moves closer to a learner’s typical level on a later attempt even without intervention. If treatment begins after the extreme score, natural reversion can be misattributed to treatment.

21. Four plausible causes changed together. The improvement may reflect one, several or their interaction. The package can be useful even when component-level attribution is uncertain.

22. Anecdotal before–after evidence. Appropriate wording: The improvement followed use of the strategy and suggests it may have helped. Treat the observation as a hypothesis generator rather than universal proof.

23. Examples: maturation, practice effects, different test difficulty, concurrent school teaching or regression to the mean. The design observes change but does not isolate its cause.

24. Unmeasured confounding. Adjustment can address variables that were measured and modelled appropriately, but important differences may remain unseen.

25. Random assignment aims to make treatment groups comparable on average before the intervention, reducing systematic confounding. Differences that emerge can therefore be attributed more credibly to the assigned condition, subject to implementation and measurement limits.

26. Different methods have different failure modes. If experimental, observational, mechanism and replication evidence converge, an explanation that survives several independent tests becomes harder to dismiss as one method’s artefact.

27. A causal explanation. After became caused, upgrading chronology into intervention-relevant causation without new evidence.

28. Association became direct causation. Drives implies reading frequency actively produces vocabulary growth and encourages an intervention inference.

29. Predicted means the earlier measure contained forecasting information. Determines suggests much stronger causal or near-inevitable control and can imply that changing or fixing the predictor fixes the outcome.

30. The randomised design may have earned a causal comparison. Rewriting it as mere association can hide the fact that changing assigned condition produced a difference under the tested conditions.

The causal independence test

A learner is developing advanced causal-language control when they can increasingly:

• separate chronology from causation;
• identify the causal burden added by because;
• distinguish because from because of grammatically without confusing grammar with evidence;
• recognise that due to sounds formal but does not strengthen proof;
• recognise temporal and causal readings of since and as;
• distinguish a world-result from an evidence-based conclusion after so or therefore;
• recognise leads to, results in, drives, produces and responsible for as causal language;
• distinguish association from measured correlation;
• distinguish prediction from intervention;
• ask what changing X would do to Y;
• draw a mechanism between proposed cause and effect;
• recognise that a plausible mechanism is not proof;
• distinguish mediator from confounder;
• generate a common-cause explanation;
• generate a reverse-causation explanation;
• recognise treatment-selection effects;
• recognise regression to the mean;
• notice when several interventions changed simultaneously;
• distinguish necessary, sufficient and contributing causes;
• prefer contributing-cause language for genuinely multi-causal outcomes;
• match anecdotal evidence to hypothesis language;
• understand the weaknesses of simple pre/post designs;
• recognise the remaining risk of unmeasured confounding in observational studies;
• understand why randomisation strengthens causal interpretation;
• use causal language without pretending randomised studies are flawless;
• value converging evidence across different methods;
• classify because as physical cause, motive/reason or evidence-to-conclusion;
• distinguish logical consequence in Mathematics from empirical causation;
• scope causal scientific claims to tested conditions;
• represent historical causation as systems of contributing factors when appropriate;
• replace personality labels with controllable mechanisms in family and educational explanations;
• detect AI upgrades from after to because;
• detect AI upgrades from association to drives or causes;
• detect AI upgrades from prediction to determines;
• detect AI downgrades that erase legitimately earned causal information;
• ask the counterfactual question “what would have happened otherwise?”;
• calibrate the final verb to the actual evidence.

The causal vocabulary field guide: fifty words that change the arrow

Causal writing becomes more accurate when learners stop treating every cause-and-effect expression as a synonym for cause.

The field guide below groups common words by the job they usually perform. The categories overlap. Context, discipline and evidence can shift the exact force. That is precisely why the words deserve auditing.

Direct production words: cause, produce, create, generate, induce

Cause. The clearest general causal verb. X caused Y says X produced a difference in Y. Use only when the evidence earns a causal interpretation.

Produce. Often causal and sometimes more mechanistic. The intervention produced a reduction in errors presents the reduction as an outcome generated by the intervention.

Create. Strong when X brings a new state into existence: The rule created a scheduling conflict. It can also be metaphorical and broad, so define the conflict rather than relying on the verb’s drama.

Generate. Common in technical writing for outputs, signals, data, heat, ideas or effects. It usually presents X as an active producer.

Induce. Strong causal language, especially in scientific or medical contexts. It often implies deliberate or reliably produced change. Avoid using it merely because it sounds technical.

Onset words: trigger, initiate, precipitate, spark

Trigger. X starts a process that may already have enabling conditions in place. A trigger need not be the deepest or only cause.

Historical example:

The incident triggered the crisis.

This can distinguish the immediate initiating event from structural conditions that made the crisis possible.

Initiate. Marks beginning. It does not automatically say X is sufficient to sustain the later process.

Precipitate. Often means to bring about an event suddenly or sooner than it otherwise might have occurred. Strong causal vocabulary and relatively formal.

Spark. Metaphorical causal language. Vivid in journalism or narrative, less precise in technical analysis unless the metaphor’s initiating role is genuinely intended.

Pathway words: lead to, result in, give rise to, culminate in

Lead to. Causal or pathway language. Often feels softer than cause, but still presents X as producing Y.

Result in. Strong directional relation from cause/process to outcome. Good when the mechanism is known or the relationship is well supported.

Give rise to. Formal causal language. It can make a sentence sound scholarly while carrying essentially the same burden as produce.

Culminate in. Marks the final outcome of a developing sequence. It tells us about endpoint and sequence, but depending on context may not identify which earlier element was causally decisive.

Contribution words: contribute to, play a role in, partly explain, factor in

Contribute to. Excellent for multi-causal outcomes. It says X has a causal role without claiming exclusivity or sufficiency.

Play a role in. Similar to contribute to but often vaguer. Ask what the role actually is.

Partly explain. Signals that X accounts for some portion or aspect of Y. Still causal or explanatory in spirit; partly does not remove the need for evidence.

Factor. Dangerous because it can mean almost anything. Motivation was a factor may be reasonable but under-specified. Prefer contributed by… or name the mechanism if possible.

Strengthening words: amplify, intensify, exacerbate, accelerate

Amplify. X makes an existing effect or signal larger. It presupposes something to amplify.

Intensify. X increases severity or strength. Useful for conflict, pressure, emotion or processes.

Exacerbate. X makes an already negative condition worse. The negative evaluation is built into the word.

Accelerate. X makes a process happen faster. This is different from causing the process to exist. A factor can accelerate learning without being the original cause of learning.

Reducing words: reduce, suppress, inhibit, mitigate, buffer

Reduce. Direct causal verb when X lowers Y. Stronger than saying X is negatively associated with Y.

Suppress. Suggests active reduction or restraint of an ongoing process.

Inhibit. X slows or prevents a process from operating normally. Common in scientific and psychological explanation.

Mitigate. X reduces severity rather than eliminating the cause or outcome entirely.

Buffer. X protects Y from another cause or stressor. This implies an interaction: the effect of one factor changes in the presence of another.

Enabling words: enable, allow, permit, facilitate, support

Enable. X makes Y possible. An enabling condition need not make Y inevitable.

Allow. Similar to enable, often less technical. The diagram allowed the learner to see the relationship can describe a mechanism without claiming the diagram alone caused mastery.

Permit. Often means removing a constraint. In causal analysis, it can distinguish “made possible” from “made happen”.

Facilitate. X makes Y easier or more likely. It is causal-ish but weaker than produce. Ask how the facilitation occurs.

Support. Highly ambiguous. It can mean evidence supports a claim, a teacher supports a learner, or one factor supports a process. The direction must be inferred from context.

Constraint words: prevent, block, constrain, limit

Prevent. Strong: X stops Y from occurring. To claim prevention, the writer needs a credible reason to believe Y would otherwise have occurred.

Block. Often mechanistic: X interrupts a pathway or access point.

Constrain. X narrows the possible range of Y rather than necessarily determining one outcome.

Limit. X places an upper bound, reduces scope or restricts possibility. More modest than prevents.

Influence words: affect, influence, impact, shape

Affect. Causal but non-specific. It says X changes Y somehow. Direction and magnitude may remain unspecified.

Influence. Also causal in ordinary use. It can sound cautious, but it still says X makes a difference.

Impact. Often vague and overused. X impacts Y does not tell us whether X increases, decreases, enables, delays or changes Y in another way. Replace with a directional verb when possible.

Shape. Broad causal-developmental language. Useful when outcomes emerge over time from many influences, but it can become metaphorical fog if no mechanism is named.

Driver language: drive, driver, determinant

Drive. Strong causal verb. Reading drives vocabulary growth is much stronger than reading is associated with vocabulary growth.

Driver. Popular in business, policy and journalism. Usually implies an important causal factor. Do not use it merely to mean “something related”.

Determinant. Context-sensitive. In some research fields it can refer broadly to factors associated with outcomes; in ordinary language it sounds strongly causal. Define what is meant rather than relying on prestige.

Explanation words: explain, account for, responsible for

Explain. Can be causal, statistical or conceptual. X explains 20% of variance in a statistical model does not necessarily mean X causes 20% of the outcome.

Account for. Similarly ambiguous. It may mean “statistically associated with part of the variation”, “provide a reason for”, or “cause”. Read the methodology.

Responsible for. Strong causal and often moral language. X is responsible for Y can assign agency or blame as well as causation. It should not be used casually in neutral analysis.

Non-causal relation words: associated with, related to, linked to, correlated with

Associated with. A good general observational relationship phrase. It does not specify causal direction.

Related to. Broad and often weaker. Useful when the exact statistical relation is not being claimed, but it can be too vague for serious analysis.

Linked to. Common in journalism. Readers may hear it as causal even when writers intend association. More explicit wording is preferable in technical work.

Correlated with. Ideally reserved for an actual correlation measure or clearly statistical relationship. Correlation by itself does not identify cause.

Prediction and evidence words: predict, forecast, indicate, suggest, signal

Predict. X improves forecasting of Y. It says nothing by itself about whether changing X changes Y.

Forecast. Similar, often used for explicit future outcomes.

Indicate. Evidence language. X points toward a conclusion but does not necessarily cause the thing indicated.

Suggest. Appropriately cautious evidence language when the pattern supports a possibility without settling it.

Signal. Can mean predictive marker, indicator or communication cue. A signal of risk is not necessarily a cause of risk.

Risk language: risk factor, increases risk, marker of risk

Risk factor. Frequently misunderstood. Depending on field and study design, a risk factor may be causal, predictive or simply associated with higher probability. Do not assume intervention on every risk factor reduces the outcome.

Increases risk. Causal-sounding and often causal in intended use. It means changing exposure increases probability of outcome under relevant conditions.

Marker of risk. Safer when X identifies higher-risk cases without evidence that X itself produces the risk.

The verb swap test: one dataset, six different claims

Constructed observation:

Students who report more sleep have higher average examination scores.

Now swap the verb:

Sleep is associated with examination performance.

Relationship only.

Sleep predicts examination performance.

Forecasting relationship, assuming prediction was actually tested.

Sleep contributes to examination performance.

Qualified causal role.

Sleep influences examination performance.

General causal effect, magnitude unspecified.

Sleep drives examination performance.

Strong important-cause framing.

Sleep determines examination performance.

Extremely strong; suggests control or near-necessity and is usually indefensible for a multi-causal outcome.

The underlying observation did not change.

The implied theory changed six times.

The field-guide rule: replace prestige with mechanism

When a causal word sounds sophisticated, ask what it means operationally.

Motivation drives achievement.

How?

Motivation facilitates sustained practice.

How?

Students who choose to remain on a difficult task for longer complete more retrieval attempts.

Now we have something observable.

The best causal vocabulary often becomes less impressive as the mechanism becomes more precise.

That is a good trade.

A parent route: replace blame with a testable explanation

Family explanations become more useful when they describe observable mechanisms instead of permanent character labels.

Instead of You were late because you are irresponsible, begin with the event:

You started packing only after the planned departure time, and the family left twelve minutes late.

Instead of You always need help, describe the pattern:

You completed four familiar questions independently but needed a first-step cue on two unfamiliar ones.

Then propose one mechanism:

The unfamiliar questions may be difficult to start because the first operation is not recognised quickly enough.

Change the suspected cause and observe the result.

Give one first-step recognition cue without solving the problem. If the child then works independently, the start-state explanation gains credibility. If nothing changes, revise the hypothesis.

The family does not need laboratory certainty to make a sensible decision. It does need enough discipline to distinguish this appears to help our child from this is the proven cause of the outcome.

A tutor route: diagnose the first broken link

Teaching is intervention, so tutors need causal models.

A wrong answer can be produced by many different upstream failures:

• unknown vocabulary;
• wrong word sense;
• reference-chain confusion;
• missed contrast;
• incorrect timeline;
• unsupported assumption;
• misread command word;
• time pressure;
• answer switching without new evidence.

Poor comprehension names the outcome. It does not identify the cause.

A useful probe sequence is:

1. Can the learner explain the question?
2. Can they locate the evidence?
3. Can they define the crucial word in context?
4. Can they track pronouns and referents?
5. Can they state the timeline?
6. Can they separate text evidence from assumption?
7. Can they answer after only the smallest missing cue is supplied?

The first failed probe suggests a bottleneck worth testing.

If supplying one word meaning repairs the answer while all later reasoning remains independent, vocabulary has a credible local causal role. If supplying the meaning changes nothing, vocabulary was probably not the first broken link.

This is why diagnosis should come before more practice. More practice strengthens the process being practised; if the causal model is wrong, it can strengthen the wrong repair.

The causal writing ladder

1. Reason and result: connect clear everyday causes and outcomes.
2. Sequence versus cause: learn that after is not because.
3. Alternative explanations: generate reverse direction and common causes.
4. Relationship type: separate association, prediction and causation.
5. Mechanism: open the arrow between X and Y.
6. Study design: match causal confidence to how evidence was produced.
7. Counterfactual: ask what would have happened otherwise.
8. Systems: recognise feedback loops and multiple contributing causes.

One late conversation in Punggol

The four students are looking at one sentence:

Her marks improved because her phone was removed during revision.

Maya says, “That makes sense.”

Hana asks, “What else changed?”

Jia Jun writes, The phone ban transformed her results.

Ethan draws an arrow:

phone outside room → fewer interruptions → longer uninterrupted study → better preparation → higher marks

“That is the mechanism we think happened,” he says.

Hana checks the rest of the month.

• bedtime moved earlier;
• timed-paper practice began;
• two difficult topics were repaired.

Maya looks at the arrow again.

“So the phone rule could still have helped?”

“Yes,” Ethan says. “We just do not know how much of the improvement belongs to it.”

Jia Jun deletes transformed.

They write:

After the phone rule was introduced, interruptions during revision fell and the student’s study periods became more continuous. Her marks also improved. The rule is therefore a plausible contributing factor, but several other changes occurred during the same period, so the available evidence does not isolate its individual effect.

Longer.

Less dramatic.

More useful.

The real goal: earn the arrow

Causal language turns a messy world into a direction.

X happened.

Y happened.

The advanced reader asks whether the evidence has earned the arrow between them.

• Did the proposed cause come first?
• Is there a mechanism?
• What would have happened otherwise?
• Could the direction run backwards?
• Could a third factor influence both?
• Were the compared groups different before the exposure or intervention?
• What design attacks those alternatives?
• What verb does that evidence deserve?

Do not weaken a real causal finding merely because caution sounds intelligent.

Do not inflate an association merely because certainty sounds impressive.

The question is not whether the causal story sounds sensible. It is how much of the arrow the evidence has earned.

Research notes and further reading

The sources below support the language, study-design and causal-inference principles. The Cause–Mechanism Audit, resident-character cases, family and tutor routes, field guide and practice sequence are original instructional material.

• Cambridge Grammar: Conjunctions — causes, reasons, results and purpose — cause, reason and result relationships using because, as, since, so and related forms.
• Cambridge Grammar: Because and because of — grammatical distinction between causal clauses and noun-phrase constructions.
• Cambridge Dictionary: Causal — general reference for causal relationships.
• NCBI Bookshelf: Principles of Epidemiology — association, comparison, study design and causal interpretation.
• NCBI Bookshelf: Causal inference and study design — confounding, observational evidence and counterfactual reasoning.
• Hernán and Robins: Causal Inference — Randomized Experiments — why random assignment changes the relationship between association and causal interpretation under ideal conditions.

Continue through the advanced eduKatePunggol vocabulary chain

• Advanced Vocabulary Sequence | The Timeline Audit
• Advanced Vocabulary Cohesion | The Reference Audit
• Advanced Vocabulary Perspective | The Framing Audit
• Advanced Vocabulary Agency | The Hidden-Agent Audit
• Advanced Vocabulary Synthesis | The Contrast Audit
• Advanced Vocabulary Argument | The Qualification Audit
• Advanced Vocabulary Reading | The Assumption Audit
• Advanced Vocabulary Editing | The Counterexample Test
• Education: Regarding Vocabulary